432
200'
MEIS-II data. The data were contrast stretched by
using a linear or log transform and displayed on the
colour video monitor. The same analysis procedure as
that used for the MEIS-II data was then followed
(Table 3).
RESULTS AND DISCUSSION
Approximately 50% of the study area was sketch mapped
to show four levels of infestation. The sketch maps
were then used as ground data during the detailed
photointerpretation of the infrared aerial photo
graphs. A 1:25,000 scale map was then produced.
Four levels of infestation: light (0-25%), moderate
(26-50%), heavy (51-75%) and severe (76-100%) were
detected and mapped using the aerial photographs.
This map acted as ground data for analysis of the
MEIS-II and TM data.
MEIS-II data
A number of contrast-stretched band combinations were
displayed to determine which gave the best indication
and separation of spruce budworm infestation, healthy
spruce, and other vegetation types at 5.5 m spatial
resolution. Band 2 (red), band 4 (green), and band 5
(blue) gave a natural colour image with the budworm
infested areas appearing reddish-brown. Some confu
sion occurred however with treed bogs, mixed white
spruce and black spruce, and some healthy white
spruce (Figure 2). The greatest confusion was
between budworm infested spruce and the vegetated
wetland areas which occupied a substantial part of
the study area. Other band combinations which gave
good visual separation were: 1) band 5 (red), band 4
(green), band 2 (blue); 2) band 4 (red), band 7
(green), and band 8 (blue); 3) band 8 (red), band 6
(green) and band 2 (blue). The analysis of various
band combinations and comparison to the aerial photo
graphy was done on a visual basis only. In drier
parts of the study area, separation of budworm infes
tation from other vegetation types was much easier.
At the visual analysis level it was only possible to
detect the severe level of infestation, although
there was some indication of the moderate level of
infestation. Areas of light infestation had a
reflectance signature similar to healthy spruce.
Separation between healthy and budworm infested
white sprue was dependent on the bands used and
training areas chosen. If great care was taken to
obtain only the purest possible reflectance values
for budworm infestation and healthy spruce, then the
overlap was minimized (Figure 2). Bands 7 and 5 gave
the least overlap (Figure 3).
When the 5.5 m resolution MEIS-II data was
resampled to 10 m, reflectance signatures for healthy
spruce overlapped part of the budworm infestation
signature (Figures 4 and 5). Again, separation was
dependent on the bands chosen and the purity of the
training areas. The lower spatial resolution also
seemed to have an effect since radiance values are
now averaged. Visual separation of the healthy and
budworm infested spruce was only possible at the
severe level. There seemed to be greater confusion
between budworm infestation, the wetland areas and
black spruce than in the original 5.5 m resolution
data. Bands 8 and 5 and bands 7 and 5 gave the best
separation (Figures 4 and 5).
The results of principal component enhancement did
not improve upon contrast stretching. Good separa
tion was obtained by using healthy spruce and budworm
infested areas as training areas. The Martin Taylor
enhancement gave good separation in some areas but
did not improve on the contrast stretched raw data.
Biomass indices were obtained for several bands.
The best separation of white spruce from other vege
tation types was obtained by displaying the biomass
index of bands 8 and 6, bands 3 and 5, and bands 7
and 2. The infested white spruce appeared as a light
Figure 2. Scattergram outlines of MEIS-II bands 2 and 8
for four classes at 5.5 m resolution.
Healthy
white
spruce
Budworm infested white spruce
0'
.66 - .69 micrometer intensity
Figure 3. Scattergran outlines of MEIS-II bands 5 and 7
for five classes at 5.5 m resolution.
Black/
spruceV,
J Aspen
\Budworm infested white spruce
]Healthy white spruce
Q Water
o 50 100 150 200 250
.66 - .69 micrometer intensity
Figure 4. Scattergram outlines of MEIS-II bands 5 and 8
for five classes at 10 m resolution.
green colour while healthy white spruce and black
spruce were darker. Separation between healthy and
infested white spruce was therefore dependent on
being able to distinguish between shades of green.
Three classification algorithms produced varying
160
Mo
to
40
0
FI
result
ficati
be us<
vised
better
puritj
same l
bud wo
classi
healti
spruce
around
classi
worm
drier
Table
using
Class
BUDW
HEAL
ASPN
BLSP
WATR
Total
Table
using
Class
BUDW
HEAL
ASPN
BLSP
WATR
Total
The
the l
drier
occur
sprue
be u
value
numbe
confu